Air-ground cooperative firebreak opening method based on digital twinning and neural network

Abstract

The application discloses a kind of based on digital twin and neural network ground-air coordination fire prevention isolation belt opening method, belong to forest and prairie fire prevention technical field.First, collect geographic information, meteorological, fire monitoring data and pre-process, construct digital twin model simulation meteorological element.Then, construct evaluation index system to evaluate fire risk, and divide risk level.Genetic algorithm is used to plan the isolation belt, and the parameter combination is explored through selection, crossover and mutation operation.When ground-air coordination is carried out, the unmanned aerial vehicle guides the construction, the helicopter hoists the materials, and the scheme is adjusted in real time.The fire resistance effect, opening efficiency and cost benefit are evaluated through simulation and actual data, and each link is optimized according to the results.The method accurately evaluates the fire risk, scientifically plans the isolation belt, improves the efficiency of ground-air coordination, realizes dynamic monitoring and optimization, reduces the cost, and provides an effective means for forest and prairie fire prevention and control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of forest and grassland fire prevention, and particularly relates to a method for opening a fire prevention isolation belt based on digital twinning and neural networks. BACKGROUND

[0002] In the global scope, forest and grassland fires occur frequently, posing a serious threat to the ecological system, economic development and human life and property safety. The traditional method of opening a fire prevention isolation belt has many defects that are difficult to overcome when faced with complex and variable forest and grassland environments.

[0003] In terms of data collection and processing, the traditional method mainly relies on manual field measurement and simple remote sensing monitoring. Manual measurement is not only inefficient and limited in coverage, but also easily affected by terrain and weather conditions, making it difficult to obtain real-time and comprehensive data. Early remote sensing technology has low resolution and cannot accurately identify the subtle features of vegetation types, density and topography, resulting in a lack of accurate data support for subsequent fire prevention isolation belt planning. In terms of data processing, the traditional method is difficult to effectively integrate and analyze a large amount of multi-source data, and cannot fully tap the potential information behind the data, making it difficult to accurately assess fire risk.

[0004] In terms of fire risk assessment, the traditional method often only considers a single or a few factors, such as vegetation type and distribution, ignoring the comprehensive influence of factors such as terrain, weather and human activities. The lack of a scientific and systematic evaluation index system makes the judgment of fire risk not accurate and comprehensive. Moreover, the traditional evaluation method is mostly based on experience and simple mathematical models, which cannot adapt to the dynamic changes of forest and grassland environments, making it difficult to accurately predict the possibility and spread trend of fire in advance, and thus unable to provide scientific basis for the planning of fire prevention isolation belts.

[0005] In the planning of isolation belts, the traditional method usually relies on experience for layout, without fully considering the spread law of fire and the risk differences of different regions. The width, length and position of the isolation belt lack scientific demonstration, resulting in some isolation belts being unable to effectively prevent the spread of fire when a fire occurs, or causing resource waste due to excessive and wide setting. In addition, the traditional planning method does not fully utilize modern information technology, making it difficult to simulate and optimize different planning schemes, and unable to adjust the planning strategy in a timely manner according to the actual situation.

[0006] During the construction process, the traditional ground-air collaborative operation mode has serious shortcomings. There is a lack of efficient communication and collaboration mechanism between ground construction equipment and air operation forces, and information transmission is not smooth, resulting in low operation efficiency. For example, when a helicopter hoists equipment and materials, it may be delayed or placed in an inaccurate position due to improper coordination with ground construction equipment, affecting the opening progress of the fireproof isolation belt. Moreover, the traditional construction method is difficult to monitor the construction progress and quality in real time, and cannot timely discover and correct the deviation in the construction process, which easily leads to uneven quality of the isolation belt and affects the fireproof effect.

[0007] With global warming and the expansion of human activities, the frequency and severity of forest and grassland fires are increasing. Extreme weather events such as high temperature, drought and strong wind increase, making forest and grassland vegetation drier and more flammable, significantly increasing fire risk. At the same time, human development activities around the forest and grassland are becoming more frequent, increasing the potential risk points of fire occurrence. Under such background, the traditional method of setting up fireproof isolation belt cannot meet the growing demand of forest fire prevention, and an innovative method integrating advanced technology is urgently needed to improve the forest fire prevention ability and reduce the loss caused by fire. SUMMARY

[0008] The purpose of the present application is to provide a ground-air collaborative fireproof isolation belt setting method based on digital twinning and neural network, to solve the problems of inaccurate data, unscientific planning, and low coordination efficiency in the traditional fireproof isolation belt setting method, to realize accurate prevention and control of forest and grassland fires, to improve the efficiency and quality of fireproof isolation belt setting, and to reduce the loss caused by fire.

[0009] To achieve the above-mentioned purpose, the present application provides a ground-air collaborative fireproof isolation belt setting method based on digital twinning and neural network, comprising the following steps:

[0010] S1, data acquisition and preprocessing: collect geographic information, meteorological and fire monitoring data, and then perform coordinate unification, format conversion, quality control and denoising operation on various data;

[0011] S2, digital twinning model construction: first, construct a geometric model that accurately restores geographic features, then add physical properties to the geometric model, and use Navier-Stokes equation to construct a meteorological model to simulate meteorological elements; finally, compare the simulation results with actual observations to calibrate model parameters;

[0012] S3, fire risk assessment: construct an evaluation index system covering vegetation, terrain, meteorology and human factors, determine the weight using AHP and perform consistency test; use neural network combined with Rothermel model to assess risk, divide low, medium and high risk levels and visualize in the digital twinning model;

[0013] S4, Isolation zone planning: Genetic algorithm is adopted for planning, a fitness function is defined to evaluate the scheme, and the chromosome is optimized through selection, crossover and mutation operations. The parameter combination of the isolation zone width is explored through binary or real number coding mutation. Finally, the planning results are displayed and the fire blocking effect is simulated and analyzed. The planning is optimized as needed;

[0014] S5, Air-ground cooperative operation: the unmanned aerial vehicle guides the ground construction equipment to plan the path and monitors the construction, and the helicopter hoists the equipment and materials and receives the command of the ground station. At the same time, the equipment operation, construction progress and quality are monitored in real time, and the scheme is adjusted according to the results, and the isolation zone planning is re-evaluated and optimized periodically;

[0015] S6, Effect evaluation and feedback: the evaluation indexes of fire blocking effect, opening efficiency and cost benefit are constructed, the actual data are evaluated through simulation and recording, and each link is optimized according to the results.

[0016] Preferably, in step S2, the geometric model is constructed according to geographic information data, a three-dimensional modeling software is used to construct a forest and grassland geometric model, the geographic features of the terrain are accurately restored, a three-dimensional reconstruction technology based on point cloud data is used to construct a tree crown model for trees, and a texture mapping method is used to model grassland.

[0017] Preferably, in step S2, a meteorological model is established, a computational fluid dynamics method is adopted, a Navier-Stokes equation is used to describe the wind field motion, real-time meteorological data is combined for solving, and the distribution and change of meteorological elements are simulated.

[0018] In the Cartesian coordinate system, the Navier-Stokes equation has the following form:

[0019] Continuity equation:

[0020]

[0021] Momentum equation:

[0022]

[0023]

[0024] Wherein, ρ is the fluid density, t is the time, u, v, w are the velocity components of the fluid in x, y, z directions, p is the pressure, μ is the dynamic viscosity, and Fx, Fy, Fz are the external force components in the directions. The wind speed and direction at different positions are obtained by solving the equation through a numerical algorithm;

[0025] The simulation results of the digital twin model are compared with the actual observation data for verification. According to the verification results, the model is calibrated and the parameters are adjusted, so that the simulation results are more consistent with the actual situation.

[0026] Preferably, in step S3, the evaluation index system is constructed including the following steps:

[0027] Considering vegetation factors, topographic factors, meteorological factors, human factors;

[0028] The analytic hierarchy process (AHP) is used to determine the weight of each evaluation index, a judgment matrix is constructed by expert scoring, the relative weight of each index is calculated, and the judgment matrix needs to be tested for consistency. The consistency index (CI) calculation formula is:

[0029]

[0030] Wherein, λ max is the largest eigenvalue of the judgment matrix, and n is the order of the judgment matrix; the random consistency index RI has a corresponding standard value according to the order n of the matrix.

[0031] Preferably, in step S3, the risk is evaluated by using a neural network combined with the Rothermel model, including the following steps:

[0032] The neural network model is used for fire risk assessment, the evaluation indexes are used as the input of the neural network, and the neural network is trained by a large amount of historical fire data and corresponding evaluation index data;

[0033] The Rothermel model is used to describe the surface fire spread speed, and the formula is:

[0034]

[0035] R is the fire spread speed;

[0036] I g is the potential energy release rate per unit area, and the calculation formula is H c is the heat value of the fuel, M is the fuel load, ρ b is the bulk density of the fuel, φ is the proportion of the effective combustion part, t d is the combustion duration;

[0037] Epsilon is an energy propagation efficiency factor related to fuel type and topographic factors; omega is a wind-assisted combustion factor, which is a function of wind speed, and is expressed as omega = 1 + a * V, where V is the wind speed and a is a coefficient related to fuel and terrain; is an environmental resistance factor, which reflects the hindering effect of terrain and slope on fire spread; by inputting different vegetation, meteorological and topographic parameters into the model, the fire spread speed and direction are predicted, providing a reference for fire risk assessment.

[0038] Preferably, in step S3, the fire risk is divided into three levels of low, medium and high, and according to the output result of the neural network model, a threshold value h is set to divide the risk level, and the threshold value h is set in the range of 0-1, and the specific steps are as follows:

[0039] Collect forest and grassland fire case data under different regions, different seasons and different climate conditions, and various environmental parameter data within a period of time before the fire occurs;

[0040] Determine the environmental factors that have a greater impact on fire risk through correlation analysis, and then use cluster analysis to preliminarily divide the fire risk level interval;

[0041] Experts in the field of forest fire prevention organize to evaluate and discuss the analysis results, comprehensively consider the actual application requirements and risk prevention and control targets, and determine the threshold value h suitable for the region;

[0042] Take the threshold value h as a lower threshold value h1 and a higher threshold value h2 respectively;

[0043] When the neural network model outputs a risk value less than h1, the region is determined to be a low risk level;

[0044] When the neural network model outputs a risk value between h1 and h2, the region is divided into a medium risk level;

[0045] When the neural network model outputs a risk value greater than h2, the region is identified as a high risk level.

[0046] Preferably, in step S4, binary coding is used to explore the isolation belt width parameter combination and fire blocking effect analysis, including the following steps:

[0047] Determine the isolation belt width represented by binary coding, and determine the actual width change corresponding to each bit of coding;

[0048] Based on the genetic algorithm, a group of initial chromosome populations containing isolation belt width parameters are randomly generated;

[0049] Mutate the isolation belt width coding in the chromosome;

[0050] Substitute the mutated chromosome into the fitness function to calculate the fitness value, and the fitness function is

[0051]

[0052] Fitness represents the fitness value, and the greater the value, the better the corresponding isolation belt planning scheme; α and β are weight coefficients; A blocked The area successfully blocked by the isolation belt simulated by the digital twin model, and the greater the value, the more significant the role of the isolation belt in preventing the spread of fire; Atotal The total area where the fire can spread without setting up the isolation belt reflects the relative effect of the isolation belt on blocking the spread of the fire; C is the total cost of setting up the isolation belt, and the lower the value of C, the more conducive to the improvement of the fitness value;

[0053] According to the fitness value, the chromosomes are selected, and the chromosomes with high fitness are retained to enter the next generation;

[0054] Through continuous iteration, the mutation, fitness calculation and selection operation are repeated, and the better isolation belt width parameter combination is gradually screened out;

[0055] The optimal isolation belt width parameter combination obtained by iteration is visualized and displayed in the digital twin model; the isolation belt is highlighted in different colors or line styles, and the position, direction and width information of the isolation belt are clearly presented, and the related attributes of the isolation belt under this width are labeled;

[0056] Using the digital twin model, the optimal width isolation belt is simulated under various fire scenarios;

[0057] By comparing the simulation results under different scenarios, the fire blocking effect of the isolation belt with the optimal width is comprehensively evaluated, and the advantages and disadvantages of the isolation belt under different environments are analyzed to provide a basis for optimization.

[0058] Preferably, the real number coding explores the isolation belt width parameter combination and the fire blocking effect analysis, including the following steps:

[0059] The real number coding is used to directly represent the width of the isolation belt;

[0060] An initial chromosome population containing isolation belt width parameters is generated by using a genetic algorithm;

[0061] Set the mutation range, and mutate the isolation belt width coding in each chromosome with a certain mutation probability;

[0062] Use the fitness function Calculate the fitness value of the mutated chromosome, simulate the spread of the fire under different width isolation belts in a specific fire scenario using the digital twin model, obtain A blocked and A total , and estimate the cost C to obtain the fitness value;

[0063] According to the fitness value, the chromosomes are selected, and the chromosomes with high fitness are retained to enter the next generation, and after multiple iterations, the chromosomes in the population gradually approach the better isolation belt width parameters;

[0064] The planning scheme corresponding to the optimal isolation belt width parameter after iteration optimization is displayed in the digital twin model;

[0065] With the help of the digital twin model, the optimal width of the isolation belt is simulated and tested in various fire scenarios, the spread state of the fire at different times is recorded, and the key indicators of the isolation belt successfully blocking the fire are counted.

[0066] By comparing the simulation data under different scenarios, the fire blocking effect of the width isolation belt under the influence of different environmental factors is analyzed in depth, providing detailed data support for subsequent scheme improvement.

[0067] Preferably, in step S5, the steps of real-time monitoring and adjustment are as follows:

[0068] The ground sensor network, unmanned aerial vehicle monitoring and helicopter-mounted equipment are used to collect data in all directions during the opening process of the fireproof isolation belt in real time.

[0069] The ground station comprehensively analyzes the collected multi-source data.

[0070] According to the data analysis and evaluation results, the operation scheme is adjusted in time, and at the same time, the operation situation after adjustment is continuously monitored.

[0071] Preferably, in step S6, the evaluation indexes include the fire blocking effect, opening efficiency and cost benefit of the isolation belt; wherein the fire blocking effect of the isolation belt is measured by the success rate of the isolation belt in preventing the spread of fire when a fire occurs, which is simulated by a digital twin model, and is specifically expressed as

[0072] The success rate = the number of times of successfully preventing the spread of fire / the total number of simulations x 100%;

[0073] The opening efficiency is the length or area of the isolation belt opened per unit time;

[0074] The cost benefit is the cost benefit ratio evaluated by comparing the cost of opening the isolation belt and the fire loss reduced due to the opening of the isolation belt.

[0075] Therefore, the ground-air cooperative fireproof isolation belt opening method based on digital twin and neural network has the following beneficial effects:

[0076] (1) The present application can accurately evaluate the fire risk, scientifically plan the position, width and length of the fireproof isolation belt and other parameters by acquiring high-precision geographic information, meteorological and fire monitoring data through multi-source data acquisition and preprocessing, and constructing a virtual environment highly similar to the real scene by using a digital twin model, combined with a neural network and a Rothermel model.

[0077] (2) The ground-air cooperative system architecture of the present application realizes the close cooperation of unmanned aerial vehicles, helicopters and ground construction equipment. It greatly improves the construction efficiency and shortens the opening time of the fireproof isolation belt.

[0078] (3) The present application realizes real-time dynamic monitoring of the opening process of the firebreak and the forest and grassland environment by using ground sensor networks, unmanned aerial vehicle monitoring and helicopter-mounted equipment. The operation plan is adjusted in a timely manner according to the monitoring results, and the firebreak is dynamically optimized.

[0079] (4) The present application avoids unnecessary resource waste and reduces the opening cost by scientifically planning the firebreak. At the same time, precise fire risk assessment and effective firebreak setting can greatly reduce the loss when a fire occurs.

[0080] (5) The present application provides intelligent support for decision-making based on neural network fire risk assessment and genetic algorithm firebreak planning, which can quickly analyze a large amount of data, automatically generate the optimal firebreak planning scheme, and dynamically adjust according to real-time data. Not only does it reduce the burden of manual decision-making, but also improves the accuracy and timeliness of decision-making.

[0081] The technical solutions of the present application will be further described below through the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0082] Figure 1 The present application is a flowchart of a method for opening a firebreak based on digital twinning and neural networks.

[0083] Figure 2 The present application is a schematic diagram of the ground-air collaborative system architecture structure of a method for opening a firebreak based on digital twinning and neural networks. DETAILED DESCRIPTION

[0084] The technical solutions of the present application will be further described below through the accompanying drawings and examples.

[0085] Unless otherwise defined, the technical terms or scientific terms used in the present application should be understood as the usual meaning understood by a person with ordinary skills in the art to which the present application belongs. The terms "first", "second" and similar words used in the present application do not represent any order, number or importance, but are used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to represent relative positional relationships, which may change accordingly when the absolute position of the described object changes.

[0086] EMBODIMENT

[0087] AsFigures 1-2 As shown, the present application provides a method for opening a ground-air cooperative fire prevention isolation belt based on digital twinning and neural networks, including the following steps:

[0088] S1, data acquisition and preprocessing: collect geographic information, weather, and fire monitoring data, and then perform coordinate unification, format conversion, quality control, and denoising operations on various types of data;

[0089] S2, digital twinning model construction: first, construct a geometric model that accurately restores geographic features, then add physical properties to the geometric model, and use the Navier-Stokes equation to construct a weather model to simulate weather elements; finally, compare the simulation results with actual observations to calibrate model parameters;

[0090] S3, fire risk assessment: construct an evaluation index system covering vegetation, terrain, weather, and human factors, determine the weights using AHP and perform consistency test; use neural networks combined with Rothermel model to assess risk, divide low, medium, and high risk levels, and visualize the results in the digital twinning model;

[0091] S4, isolation belt planning: use genetic algorithm planning, define fitness function to evaluate the scheme, optimize the chromosome through selection, crossover, and mutation operations, explore the parameter combination of isolation belt width through binary or real number coding mutation, and finally display the planning results and simulate the fire blocking effect for optimization as needed;

[0092] S5, ground-air cooperative operation: unmanned aerial vehicles guide ground construction equipment to plan paths and monitor construction, helicopters transport equipment and materials and receive ground station commands; at the same time, monitor equipment operation, construction progress and quality in real time, adjust the scheme based on the results, and periodically re-evaluate and optimize the isolation belt planning;

[0093] S6, effect evaluation and feedback: construct evaluation indexes for fire blocking effect, opening efficiency, and cost-benefit, evaluate through simulation and actual data recording, and feedback and optimize each link according to the results.

[0094] As shown, Figure 2 The ground-air cooperative system architecture includes ground monitoring, execution subsystem, and air monitoring and support subsystem, wherein:

[0095] The ground monitoring and execution subsystem is as follows:

[0096] Sensor network: deploy a large number of sensor nodes in forest and grassland areas, including temperature, humidity, smoke, infrared sensors, etc., and transmit collected data to the ground base station in real time through wireless ad hoc network.

[0097] Ground station: receives sensor network data, performs preliminary processing and analysis. Using digital twin models, combined with real-time data, the current state of the forest and grassland is visualized, including vegetation distribution, weather conditions, etc. At the same time, according to the fire risk assessment algorithm, the fire risk level of different regions is calculated.

[0098] Ground construction equipment: including bulldozers, brush cutters and other mechanical equipment for creating firebreaks, remotely controlled by the ground station according to decision instructions or operated by operators according to the planned path.

[0099] Air monitoring and support subsystem, as follows:

[0100] Unmanned aerial vehicle monitoring: equipped with multiple unmanned aerial vehicles, equipped with high-definition cameras, thermal imagers, laser radars, etc. Patrol according to the preset route, real-time image and video transmission back to the ground station. Use neural network algorithm to analyze images and quickly find potential fire sources and fire signs.

[0101] Helicopter support: when a fire occurs, a helicopter quickly arrives at the scene. It can carry fire-fighting supplies to fight fires in the air, and hoist large ground construction equipment to areas that are difficult to reach to assist in creating firebreaks. The helicopter is equipped with communication equipment and communicates with the ground station in real time to receive command instructions.

[0102] In step S1, data collection and preprocessing includes geographic information data collection, weather data collection, fire monitoring data collection, and data preprocessing, wherein:

[0103] Geographic information data collection, as follows:

[0104] Use satellite remote sensing to obtain large-area regional topographic data with a resolution of 10-30 meters, extract vegetation types, land cover, etc. through image interpretation;

[0105] Use laser radar for high-precision topographic mapping to generate centimeter-level digital elevation models (DEM) to accurately reflect terrain undulations;

[0106] Unmanned aerial vehicle low-altitude mapping to obtain high-resolution images and terrain data in local areas to supplement satellite remote sensing and laser radar data.

[0107] Weather data collection, as follows:

[0108] Multiple weather monitoring stations are set up in the region and its surroundings to monitor real-time weather parameters such as wind speed, wind direction, temperature, humidity, and air pressure, which are transmitted to the ground station through a wireless communication network;

[0109] Use weather satellite data to obtain regional macro weather information, such as weather system movement trends and precipitation distribution, and integrate it with ground weather monitoring station data to improve accuracy and comprehensiveness.

[0110] Fire monitoring data collection, as follows:

[0111] Using ground smoke, infrared sensor network, real-time monitoring of early signs of fire, such as smoke concentration, temperature anomaly, etc.

[0112] Using satellite thermal infrared sensors to monitor fire hot spots in large areas, determine the approximate location and scope of the fire.

[0113] Unmanned aerial vehicle carrying thermal imager for suspected fire area close-range, high-precision monitoring, accurate identification of fire source location and fire size.

[0114] Data preprocessing, as follows:

[0115] Coordinate unification, format conversion and other processing of geographic information data, so that it can be integrated in the digital twin model;

[0116] Quality control of meteorological data, elimination of outliers and erroneous data, use of Kriging interpolation method to supplement missing data, ensure continuity;

[0117] Denoising of fire monitoring data, use of image enhancement algorithm to process unmanned aerial vehicle shot fire images, highlight fire source features.

[0118] In step S2, the digital twin model construction includes geometric model construction, physical model construction and model verification and calibration, wherein:

[0119] Geometric model construction:

[0120] According to the geographic information data, use three-dimensional modeling software to build forest and grassland geometric model, accurately restore topography, mountains, rivers, lakes and other geographical features. For vegetation, use different modeling methods according to different types, such as tree crown model built by three-dimensional reconstruction technology based on point cloud data for trees, and texture mapping method for grassland.

[0121] Physical model construction:

[0122] Add physical properties to the digital twin model, including vegetation burning characteristics (ignition point, heat value, burning speed, etc.), soil thermal conductivity characteristics, water evaporation characteristics, etc., determined by experimental data and theoretical calculation;

[0123] Establish a meteorological model, use computational fluid dynamics (CFD) method, use Navier-Stokes equation to describe wind field motion, combine real-time meteorological data to solve, simulate wind field, temperature field, humidity field and other meteorological element distribution and change. In the Cartesian coordinate system, the form of Navier-Stokes equation is as follows:

[0124] Continuity equation:

[0125]

[0126] Momentum equation:

[0127]

[0128] where p is the fluid density, t is time, u, v, w are the velocity components of the fluid in the x, y, z directions, p is the pressure, and m is the dynamic viscosity, respectively, are the external force components in the direction. The wind speed and direction at different positions are obtained by solving the equation by numerical algorithm;

[0129] The simulation results of the digital twin model are compared and verified with the actual observation data. According to the verification results, the model is calibrated and the parameters are adjusted to make the simulation results more consistent with the actual situation. If there is a deviation between the simulated wind speed and the measured wind speed, adjust the roughness parameters in the weather model until the error is within an acceptable range.

[0130] Model verification and calibration:

[0131] Consider vegetation factors (vegetation type, density, water content), terrain factors (slope, aspect, elevation), meteorological factors (wind speed, wind direction, temperature, humidity), and human factors (human activity density, fire source distribution).

[0132] The analytic hierarchy process (AHP) is used to determine the weight of each evaluation index. The judgment matrix is constructed by expert scoring, and the relative weight of each index is calculated. The judgment matrix needs to be tested for consistency, and the consistency index CI is calculated as follows:

[0133]

[0134] where l is the largest eigenvalue of the judgment matrix, and n is the order of the judgment matrix; the random consistency index RI has a corresponding standard value according to the order n of the matrix. max

[0135] The consistency ratio CR is expressed as

[0136]

[0137] When CR < 1, the judgment matrix is considered to have acceptable consistency, otherwise it needs to be adjusted.

[0138] In step S3, the neural network is used in combination with the Rothermel model to evaluate the risk, including the following steps:

[0139] The neural network model is used for fire risk assessment, and the evaluation indexes are used as the input of the neural network. The neural network is trained by a large amount of historical fire data and corresponding evaluation index data;

[0140] ​The Rothermel model is used to describe the surface fire spread speed, and the formula is as follows:

[0141]

[0142] R is the fire spread speed (m / min);

[0143] I g is the potential energy release rate per unit area (kW / m 2 ), and the calculation formula is H c is the heat value of fuel (kJ / kg), M is the fuel load (kg / m 2 ), ρ b is the bulk density of fuel (kg / m 3 ), φ is the proportion of effective combustion, t d is the combustion duration (min);

[0144] ε is the energy propagation efficiency factor, which is related to fuel type and terrain factors; ω is the wind-assisted combustion factor, which is a function of wind speed, and is expressed as ω = 1 + a·V, where V is the wind speed and a is a coefficient related to fuel and terrain; is the environmental resistance factor, which reflects the hindering effect of terrain and slope on fire spread; by inputting different vegetation, meteorological and terrain parameters into the model, the fire spread speed and direction are predicted, providing a reference for fire risk assessment.

[0145] The fire risk is divided into three levels: low, medium and high. According to the output results of the neural network model, the threshold value h is set to divide the risk levels, and the value range of the threshold value h is set between 0-1, and the specific steps are as follows:

[0146] Collect forest and grassland fire case data under different regions, different seasons and different climate conditions, as well as various environmental parameter data within a period of time before the fire occurs;

[0147] Determine the environmental factors that have a greater impact on fire risk through correlation analysis, and then use cluster analysis to preliminarily divide the fire risk level interval;

[0148] Organize experts in the field of forest fire prevention to evaluate and discuss the analysis results, and comprehensively consider the actual application requirements and risk prevention and control targets to determine the threshold value h suitable for the region;

[0149] Take the threshold value h as a lower threshold value h1 and a higher threshold value h2;

[0150] When the output risk value of the neural network model is less than h1, the region is determined to be at a low risk level;

[0151] When the output risk value of the neural network model is between h1 and h2, the region is divided into a medium risk level;

[0152] When the risk value output by the neural network model is greater than h2, the region is identified as a high-risk area.

[0153] In step S4, the isolation zone planning includes planning principles, a planning method based on optimization algorithms, and the presentation and analysis of planning results, wherein:

[0154] Planning Principles: Adhering to the principles of "fire prevention, fire control, and ease of establishment." Firebreaks should be established in locations that effectively prevent the spread of fire, such as along ridgelines, rivers, roads, and other natural or artificial barriers. Construction convenience should also be considered, selecting areas with flat terrain and relatively sparse vegetation. Digital twin model information should be used to comprehensively consider factors such as fire risk level, topography, and vegetation distribution. Firebreaks should be prioritized and established in high-risk areas, with appropriately increased width and density.

[0155] Planning methods based on optimization algorithms:

[0156] A genetic algorithm (GA) is used for segregation zone planning. The location, length, width, and other parameters of the segregation zone are encoded into chromosomes.

[0157] The fitness function is defined to evaluate the quality of a planning scheme. Its fitness function is:

[0158]

[0159] Fitness represents the fitness value; the higher the value, the better the corresponding buffer zone planning scheme. α and β are weighting coefficients. blocked The larger the value of the firebreak simulated by the digital twin model, the more significant the firebreak's effect in stopping the fire; A total The fire could spread over the total area without firebreaks, reflecting the relative effectiveness of firebreaks in preventing the spread of fire; C is the total cost of setting up firebreaks. The lower the C value, the more beneficial it is to improving the fitness value.

[0160] The chromosome is iteratively optimized through genetic operations of selection, crossover, and mutation. The chromosome code corresponds to the planning parameters of the isolation band, and the genes in the chromosome represent the position, length, and width information of the isolation band.

[0161] Assume the parent chromosome P1 = [a1, a2, ..., a n ] and P2 = [b1,b2,…,b n ], where a i and b i , i = 1, 2, ..., n, are specific parameter values ​​related to the planning of the isolation zone;

[0162] Randomly select a crossover point k, 1 < k < n, then the two offspring chromosomes C1 and C2 generated after crossover are:

[0163] C1 = [a1, a2, …, a k ,b k+1 ,b k+2 ,…,b n ];

[0164] C2 = [b1, b2, …, b k ,a k+1 ,a k+2 ,…,a n ];

[0165] Mutation operation can introduce new genetic characteristics. For genes representing the width of the isolation belt, if binary encoding is used, the corresponding encoding in the chromosome is "0010", which represents an isolation belt width of 8 meters (assuming that each binary encoding corresponds to a width change of 2 meters). In actual fire simulation scenarios, an 8-meter-wide isolation belt may not be sufficient in high wind and flammable vegetation conditions, as part of the fire may break through the isolation belt, indicating that the width setting is conservative. In the mutation operation, if the third "1" becomes "0", the encoding becomes "0000", corresponding to an isolation belt width of 0 meters, which does not meet the actual requirements and has a low fitness value. If the fourth "0" becomes "1", the encoding becomes "0011", corresponding to an isolation belt width of 10 meters. After re-simulation, a 10-meter-wide isolation belt successfully blocks the spread of fire, with a small increase in construction cost, far lower than the loss of fire out of control, and the fitness value is significantly improved, exploring a more suitable width.

[0166] If real number encoding is used, assume the original isolation belt width gene value is 10 meters, and set the mutation range to be 20% up and down from the original value. In the mutation operation, a mutation coefficient between -0.2 and 0.2 is randomly generated, and suppose the generated mutation coefficient is 0.15, then the mutated isolation belt width gene value is 10 × (1 + 0.15) = 11.5 meters.

[0167] Substitute this new width value into the isolation belt planning scheme and simulate the fire scenario using the digital twin model. The results show that an 11.5-meter-wide isolation belt not only effectively blocks the fire, but also has a smaller increase in construction cost compared to a 10-meter-wide isolation belt, further improving the fire-blocking effect. After calculating the fitness function, it is found that the fitness value of the mutated scheme is higher than that of the original scheme, verifying that through mutation operation under real number encoding, a more optimal isolation belt width parameter is successfully found, achieving effective fire blocking and reasonable cost control, and improving the overall optimization effect of the planning scheme.

[0168] Planning result display and analysis:

[0169] The optimized isolation belt planning scheme is visualized in the digital twin model, which intuitively presents the position, direction, width, and other information of the isolation belt. Different levels of isolation belts are distinguished by different colors or line styles.

[0170] The digital twin model is used to simulate and analyze the planning scheme, and to evaluate the fire-blocking effect of the isolation belt under different fire scenarios. For example, under different wind speed and direction conditions, the spread of fire with and without isolation belts is simulated, and the inhibitory effect of the isolation belt on fire spread is compared and analyzed. Based on the simulation analysis results, the planning scheme is further optimized and adjusted.

[0171] In step S5, the ground-air cooperative operation is as follows:

[0172] The unmanned aerial vehicle guides the ground construction equipment:

[0173] The unmanned aerial vehicle flies above the planned isolation belt area, takes real-time images and transmits them back to the ground station. Image processing algorithms are used to identify ground terrain features and obstacles, and to plan a safe and efficient driving path for the ground construction equipment.

[0174] The ground construction equipment receives the path information transmitted by the unmanned aerial vehicle and drives according to the path through an automatic driving system or an operator. The unmanned aerial vehicle continuously monitors the construction progress and quality, and timely detects and corrects construction deviations.

[0175] Helicopter hoisting equipment and materials:

[0176] In areas with complex terrain where ground construction equipment cannot reach, the helicopter hoists small construction equipment to the designated location, and also hoists fire-fighting materials to the vicinity of the fire site.

[0177] During the helicopter hoisting process, real-time communication is carried out with the ground station to receive flight instructions and safety prompts. The ground station plans a safe flight route for the helicopter based on the digital twin model terrain information and weather data, avoiding obstacles and adverse weather areas.

[0178] Real-time monitoring and adjustment:

[0179] The ground sensor network, unmanned aerial vehicle monitoring, and helicopter-mounted equipment are used to monitor the progress and quality of the fire prevention isolation belt in real time.

[0180] Based on the real-time monitoring results, the operation scheme is adjusted. If the construction progress is lagging, additional construction equipment or adjusted construction process is added; if the quality of the isolation belt does not meet the standards, timely rework is performed.

[0181] During real-time monitoring, the running state of the equipment also needs to be paid attention to. By installing sensors on the construction equipment and unmanned aerial vehicles, helicopters, etc., data such as vibration, temperature, and oil consumption of the equipment are collected, and a machine learning algorithm is used to establish a device fault prediction model. Once abnormal signs are found in the equipment, maintenance is arranged in advance to avoid equipment failure during operation and affect the opening progress of the firebreak.

[0182] At the same time, considering the dynamic changes of forest and grassland environment, every certain time period, the latest geographic information data, meteorological data and fire monitoring data are used to re-run the fire risk assessment model and the isolation belt planning algorithm, and the existing isolation belt is adjusted and improved according to the evaluation results, such as widening the isolation belt with insufficient width, and re-planning and opening the isolation belt with unreasonable position.

[0183] In step S6, the effect evaluation and feedback are as follows:

[0184] Evaluation index:

[0185] The fire resistance effect of the isolation belt is measured by the success rate of the isolation belt in stopping the spread of fire when the fire occurs. Success rate = number of times of successfully stopping the spread of fire / total number of simulations x 100%. At the same time, the average fire resistance time of the isolation belt under different fire scenarios can also be evaluated, that is, the average length of time from the occurrence of the fire to the complete prevention of the fire by the isolation belt.

[0186] Efficiency of opening, calculate the length or area of the isolation belt opened per unit time. In addition, considering the use efficiency of the equipment, the proportion of actual operation time to total operation time of the construction equipment can also be calculated to evaluate the utilization efficiency of equipment resources. If the proportion is low, it means that there is a problem of equipment idling or too long waiting time in the construction process, which needs to be further optimized.

[0187] Cost-benefit, compare the cost of opening the isolation belt with the fire loss (including forest resource loss, ecological environment damage loss, fire extinguishing cost, etc.) reduced due to the opening of the isolation belt, and evaluate the cost-benefit ratio. For forest resource loss, it can be estimated according to different tree species, forest age and area, combined with market timber price and ecological service value evaluation method; the ecological environment damage loss considers the long-term impact of fire on soil, water, biodiversity, etc., and is quantified through an ecological economics model.

[0188] Evaluation method:

[0189] Use the digital twin model to simulate multiple fires, and evaluate the fire resistance effect of the isolation belt under different fire scenarios. Each simulation records the path, speed of fire spread and isolation belt blocking in detail, and the success rate of fire resistance and average fire resistance time of the isolation belt are obtained through statistical analysis of a large amount of simulation data.

[0190] Record the construction time, equipment usage and other data during the actual construction process to calculate the opening efficiency. Through the GPS positioning system and the working state monitoring equipment installed on the construction equipment, the working track, working time and working amount of each stage of the equipment are recorded in real time, so that the length or area of the isolation belt opened in unit time and the equipment usage efficiency are accurately calculated.

[0191] By counting the loss data after the fire occurs, combined with the isolation belt opening cost data, the cost benefit ratio is calculated. Cooperate with the forestry department, ecological environment monitoring agency and the like to obtain the damaged evaluation report of the forest resources and ecological environment after the fire, and meanwhile, the cost expenditure details in the construction process are sorted out to conduct comprehensive cost benefit analysis.

[0192] Feedback and optimization:

[0193] According to the effect evaluation result, the problems existing in the system are fed back. If the fire blocking effect of the isolation belt is not ideal, the reasons are analyzed and the planning and construction links are optimized according to the problems.

[0194] If the opening efficiency is low, whether there is optimization space in the construction process, whether the construction equipment needs to be replaced and the like are analyzed.

[0195] If the cost benefit ratio is not high, the measures for reducing the cost or improving the benefit are researched.

[0196] Therefore, the fire isolation belt opening method based on digital twinning and neural network is adopted to solve the problems of inaccurate data, unscientific planning and low coordination efficiency existing in the traditional fire isolation belt opening method, realize the accurate prevention and control of the forest and grassland fire, improve the efficiency and quality of the fire isolation belt opening, and reduce the loss caused by the fire.

[0197] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application but not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can still be modified or replaced by the equivalent, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for establishing a ground-air coordinated firebreak based on digital twins and neural networks, characterized in that: Includes the following steps: S1. Data Acquisition and Preprocessing: Collect geographic information, meteorological, and fire monitoring data, and then perform coordinate unification, format conversion, quality control, and noise reduction on various types of data; S2. Digital Twin Model Construction: First, construct a geometric model that accurately restores geographical features. Then, add physical attributes to the geometric model and use the Navier-Stokes equations to construct a meteorological model to simulate meteorological elements. Finally, compare the simulation results with actual observations and calibrate the model parameters. In step S2, the geometric model is constructed based on geographic information data and using 3D modeling software to construct a geometric model of forest and grassland, accurately restoring the geographical features of the terrain. Tree canopy models are constructed using 3D reconstruction technology based on point cloud data, and grassland models are constructed using texture mapping. In step S2, a meteorological model is established, and the wind field motion is described using the Navier-Stokes equations by computational fluid dynamics. The model is then solved using real-time meteorological data to simulate the distribution and changes of meteorological elements. The simulation results of the digital twin model are compared and verified with actual observation data. Based on the verification results, the model is calibrated and the parameters are adjusted to make the simulation results more consistent with the actual situation. S3. Fire Risk Assessment: Construct an assessment index system covering vegetation, topography, meteorology, and human factors, use AHP to determine weights and conduct consistency checks; use neural networks combined with the Rothermel model to assess risks, classify them into low, medium, and high risk levels, and visualize them in a digital twin model; In step S3, the risk is assessed using a neural network combined with the Rothermel model, including the following steps: Fire risk assessment is carried out using a neural network model. The assessment indicators are used as input to the neural network, and the neural network is trained with a large amount of historical fire data and corresponding assessment indicator data. The Rothermel model is used to describe the surface fire spread rate, and the formula is: ； The speed at which the fire spreads; The potential energy release rate per unit area is calculated using the following formula: , It is the calorific value of the fuel. It is fuel load. It is the bulk density of the fuel. It is the proportion of the effective combustion portion. It is the duration of combustion; It is the energy transfer efficiency factor, which is related to fuel type and terrain factors; It is the wind-assisted combustion factor, a function of wind speed, expressed as: ,in V It's wind speed. It is a coefficient related to fuel and terrain; it is an environmental resistance factor that reflects the hindering effect of terrain and slope on fire spread; by inputting different vegetation, meteorological and terrain parameters into the model, it can predict the speed and direction of fire spread and provide a reference for fire risk assessment. S4. Firebreak Planning: Genetic algorithm is used for planning. A fitness function is defined to evaluate the scheme. Chromosomes are optimized through selection, crossover, and mutation operations. The parameter combination of the firebreak width is explored through binary or real number encoding mutation. Finally, the planning results are displayed and its fire-stopping effect is simulated and analyzed. Optimization is performed as needed. S5. Ground-air collaborative operation: UAVs guide ground construction equipment to plan the path and monitor the construction, while helicopters transport equipment and materials and receive commands from the ground workstation; at the same time, the operation of equipment, construction progress and quality are monitored in real time, the plan is adjusted according to the results, and the isolation zone planning is re-evaluated and optimized regularly. S6. Effect Evaluation and Feedback: Construct evaluation indicators for fire-retardant effect, opening efficiency, and cost-effectiveness. Evaluate through simulation and recording of actual data, and optimize each link based on the feedback of the results.

2. The method for establishing a ground-air coordinated firebreak based on digital twins and neural networks according to claim 1, characterized in that: In step S3, constructing the evaluation index system includes the following steps: Consider vegetation factors, topographic factors, meteorological factors, and human factors; The Analytic Hierarchy Process (AHP) is used to determine the weights of each evaluation indicator. A judgment matrix is ​​constructed through expert scoring, and the relative weights of each indicator are calculated. The judgment matrix needs to undergo a consistency test, and consistency indicators are used to verify the consistency of the indicators. The calculation formula is: ； in, It is the largest eigenvalue of the matrix. It determines the order of the matrix; the random consistency index. RI Based on matrix order There are corresponding standard values.

3. The method for establishing a ground-air coordinated firebreak based on digital twins and neural networks according to claim 1, characterized in that: In step S3, the fire risk is divided into three levels: low, medium, and high. Based on the output of the neural network model, a threshold h is set to classify the risk level. The value of the threshold h is set between 0 and 1. The specific steps are as follows: Collect data on forest and grassland fire cases in different regions, seasons, and climatic conditions, as well as various environmental parameter data in the period before the fire occurred; Correlation analysis was used to identify environmental factors that have a significant impact on fire risk, and then cluster analysis was used to preliminarily divide the fire risk level ranges. Experts in the field of forest fire prevention were organized to evaluate and discuss the analysis results, and, taking into account practical application needs and risk prevention and control objectives, determine a suitable threshold h for the region. Extract the lower threshold h1 and the higher threshold h2 respectively; When the risk value output by the neural network model is less than h1, the region is determined to be at a low risk level. When the risk value output by the neural network model is between h1 and h2, the region is classified as a medium-risk area. When the risk value output by the neural network model is greater than h2, the area is identified as a high-risk area.

4. The method for establishing a ground-air coordinated firebreak based on digital twins and neural networks according to claim 1, characterized in that: In step S4, the binary encoding exploration of the combination of fire-retardant parameters for the isolation zone width and the analysis of its fire-retardant effect include the following steps: Determine to use binary encoding to represent the width of the segregation band, and clarify the actual width change corresponding to each bit of the encoding; Based on a genetic algorithm, an initial chromosome population containing parameters such as the width of the isolation band is randomly generated; Mutate the encoding of the width of the segregating band in the chromosome; The mutated chromosome is substituted into the fitness function to calculate the fitness value. The fitness function is: ； This represents the fitness value; the larger the value, the better the corresponding buffer zone planning scheme. , These are weighting coefficients; The larger the value of the firebreak simulated by the digital twin model, the more significant the role of the firebreak in preventing the spread of fire. This represents the total area over which the fire could potentially spread without firebreaks, reflecting the relative effectiveness of firebreaks in preventing the spread of fire. C The total cost of setting up the isolation zone, C The lower the value, the more beneficial it is to improving the fitness value; Chromosomes are selected based on their fitness values, and those with high fitness are retained for the next generation. Through continuous iteration, repeated mutation, fitness calculation and selection operations, better combinations of isolation band width parameters are gradually selected. The digital twin model visualizes the planning scheme corresponding to the optimal combination of isolation zone width parameters obtained through iteration; the isolation zone is highlighted with different colors or line styles, clearly presenting its location, direction and width information, and marking the relevant attributes of the isolation zone under that width; Using digital twin models, the optimal width of firebreaks is simulated under various fire scenarios; By comparing simulation results under different scenarios, the fire-stopping effect of this width of firebreak is comprehensively evaluated, and its advantages and disadvantages in different environments are analyzed to provide a basis for optimization.

5. The method for establishing a ground-air coordinated firebreak based on digital twins and neural networks according to claim 1, characterized in that: The analysis of the combination of fire-retardant parameters for the isolation zone width using real-number coding includes the following steps: The width of the isolation band is directly represented by real number encoding; An initial chromosome population containing parameters such as the width of the isolation band is generated using a genetic algorithm; Define the mutation range, encode the width of the segregation band in each chromosome, and perform mutation with a certain mutation probability; Using the fitness function The fitness values ​​of the mutated chromosomes were calculated, and a digital twin model was used to simulate the fire spread of firebreaks of different widths in a specific fire scenario to obtain... and Combined with cost estimation, the following was obtained. This allows us to derive the fitness value. Chromosome selection is performed based on fitness values, and chromosomes with high fitness are retained for the next generation. Through multiple iterations, the chromosomes in the population continuously move closer to the optimal isolation band width parameter. The planning scheme corresponding to the optimal isolation strip width parameter after iterative optimization is displayed in the digital twin model; Using a digital twin model, the optimal width of the firebreak was simulated and tested in various fire scenarios. The spread of the fire at different times was recorded, and the key indicators of the firebreak's success in blocking the fire were statistically analyzed. By comparing simulation data under different scenarios, we can conduct an in-depth analysis of the fire-resistant effect of this width of firebreak under the influence of different environmental factors, providing detailed data support for subsequent scheme improvements.

6. The method for establishing a ground-air coordinated firebreak based on digital twins and neural networks according to claim 1, characterized in that: In step S5, the real-time monitoring and adjustment steps are as follows: By utilizing ground sensor networks, drone monitoring, and helicopter-borne equipment, data on the establishment of firebreaks can be collected in real time from all directions. The ground workstation performs comprehensive analysis on the collected multi-source data; Based on the data analysis and evaluation results, the work plan should be adjusted in a timely manner, and the work situation after the adjustment should be continuously monitored.

7. The method for establishing a ground-air coordinated firebreak based on digital twins and neural networks according to claim 1, characterized in that: In step S6, the evaluation indicators include the fire-stopping effect, opening efficiency, and cost-effectiveness of the firebreak; among which, the fire-stopping effect of the firebreak is measured by the success rate of the firebreak in preventing the spread of fire when a fire occurs, as simulated by a digital twin model, specifically expressed as follows: Success rate = (Number of times the fire was successfully stopped / Total number of simulations) × 100%; The opening efficiency is calculated as the length or area of ​​the isolation zone opened per unit of time. Cost-effectiveness is assessed by comparing the cost of creating firebreaks with the reduction in fire damage resulting from their creation.

Images